A semiconductor memory is provided which is operable in at least a test mode. Such semiconductor memory includes a memory array, the memory array including a plurality of memory cells which are accessible through a plurality of respective wordlines and a plurality of respective bitlines. A row decoder driver is operable to activate one wordline of the plurality of wordlines at a first point in time that is determined in relation to a first signal and to deactivate the wordline at a second point in time that is determined in relation to a second signal. The semiconductor memory further includes a precharge circuit which is operable to precharge the bitline at a third point in time, the third point in time occurring no sooner than the second point in time. A variable length delay circuit is operable to output the second signal at a delayed timing after the first signal and delayed in relation to the first signal, the delayed timing having a controllably variable length.
|
1. A method of varying a length of an active duty cycle of a wordline of a semiconductor memory, comprising:
setting a variable autoprecharge timing to a first value;
activating the wordline at a first point in time; and
autoprecharging the wordline at a second point in time determined at least partly by the first value, the length of the wordline active duty cycle defined by the length of time between the first and second points in time.
10. A semiconductor memory operable in at least a test mode, comprising:
a memory array including a plurality of memory cells accessible through a plurality of respective wordlines and a plurality of respective bitlines;
a row decoder driver operable to activate one wordline of the plurality of wordlines at a first point in time determined in relation to a first signal and to deactivate the wordline at a second point in time determined in relation to a second signal;
precharge circuit operable to precharge one of the plurality of bitlines at a third point in time, the third point in time occurring no sooner than the second point in time; and
a variable length delay circuit operable to output the second signal at a delayed timing, delayed in relation to the first signal and occurring after the first signal, the delayed timing having a controllably variable length.
6. A method of testing a memory cell within a semiconductor memory having a test mode interface, comprising:
activating the test mode interface of the semiconductor memory;
entering information through the test mode interface to set a variable autoprecharge timing measured between a start of an activation interval of a wordline and a start of a precharge interval to a first value;
switching a bitline connected to the memory cell to a first logic level;
providing an external precharge signal to the semiconductor memory;
activating the wordline to write the memory cell according to the first logic level, wherein an ending point of the active duty cycle of the wordline is determined at least partly by the external precharge signal;
providing a signal to select the autoprecharge timing;
switching the bitline to a second logic level;
activating the wordline to write the memory cell according to the second logic level, wherein the length of the wordline activation interval is determined by the first value; and
reading the contents of the memory cell to determine whether the memory cell successfully retains the second logic level.
2. The method as claimed in
3. The method as claimed in claimed in
4. The method as claimed in
5. The method as claimed in
7. The method as claimed in
8. The method as claimed in
9. The method as claimed in
11. The semiconductor memory as claimed in
12. The semiconductor memory as claimed in
13. The semiconductor memory as claimed in claimed in
14. The semiconductor memory as claimed in claimed in
15. The semiconductor memory as claimed in
16. The semiconductor memory as claimed in
17. The semiconductor memory as claimed in
18. The semiconductor memory as claimed in
19. The semiconductor memory as claimed in
20. The semiconductor memory as claimed in
|
The present invention relates to semiconductor integrated circuits and their testing.
Production testing of semiconductor integrated circuits, also referred to as “chips” herein, requires that not only devices which fail solidly be identified and dealt with, but also devices which operate marginally, i.e., those whose operation is near failing. It is common for marginal devices to be taken offline and replaced with devices which test good in order to avoid them failing later when a more costly repair would be required. Alternatively, marginal devices can be allowed to remain a part of the operating configuration, provided that the chip is appropriately tagged and the marginality of its devices are identified to downstream users which incorporate the marginal chip into further equipment. Stated another way, some chips which operate satisfactorily under certain restricted conditions, e.g., less frequent use, shorter lifetime, shorter power-on-time per day, or under better controlled temperatures, but which are likely to fail outside of such conditions may nevertheless be usable if tagged as being marginal and restricted to such use or uses.
Semiconductor memories include particular devices that are subject to variation in operating characteristics, such as due to unavoidable non-uniformities in the processing of the devices. This is particularly true of dynamic random access memories (“DRAMs”), because the design of DRAMS tends to favor the optimization of circuit density and cost. Such variations can cause some of the millions (or even billions) of memory cells of a DRAM to either fail at the time of production test, or to operate marginally at that time. Unfortunately, to downstream equipment which incorporates the chip, marginally operating memory cells can inflict greater harm than those which fail solidly during initial testing. Frequently, the marginally operating memory cell will fail hard later during the expected lifetime of the DRAM. However, today's expectations of long-term equipment reliability and performance are adverse to memory cell failures which occur later during the useful lifetime of the equipment. Even if only one of the marginal memory cells of the DRAM fails later on, reliability and performance are impacted in ways that are not acceptable to the end user.
Therefore, to assure the long-term reliability and usability of the chip, marginally operating memory cells as well as failing memory cells must be removed from the configuration at the time of production testing. Marginally operating memory cells and failing memory cells are typically replaced with redundancy memory cells available on the chip. In this way, the DRAM remains addressable by devices external to the DRAM as one intact unit offering storage and retrieval of data over one continuous address space.
On the other hand, sometimes a large number of the memory cells of a DRAM chip exhibit the same degree of marginality. If the degree of marginality is not too extreme, for example, the refresh interval must be set to 64 ms instead of 128 ms, such marginality may be deemed to be acceptable when the particular chip is tagged and used for a particular purpose that is not considered to be “mission critical.” In such case, the large number of marginal memory cells make redundancy replacement impractical or impossible; however, the degree of the marginality is considered acceptable for the particular purpose for which the chip is used. In a third case, both the number of marginal memory cells and their degree of marginality are so great that it is impractical to repair or use the chip as a marginally performing chip. In such case, the chip is identified to be scrapped.
Of course, the problem is how to identify and characterize marginally operating memory cells such that an informed decision can be made whether the marginal memory cells should be repaired and replaced with tested good redundancy cells, instead of being tagged and sold as a marginal chip, or otherwise scrapped.
It has been estimated that a single-bit failure is the most prevalent type of defect in a DRAM. Redundancy replacement designs of DRAMs typically allow a prevalence of single bit failures of up to about 1%. Single-bit failures frequently result from variations in the manufacturing process which affect either the capacitance of the storage capacitor element of the DRAM cell, or cause leakage current off the DRAM cell to exceed limits. While such failures are frequently manifested during production testing under standard operating conditions, more severe test conditions are required to investigate marginal DRAM cells.
Many DRAMs include memory cells or “DRAM cells” which have only one capacitor and one transistor, the transistor being used as a passgate device or “pass-gate transistor” for controlling the reading and writing of the DRAM cell.
In a DRAM cell a datum is stored as a charge on the storage capacitor. The pass-gate transistor is turned on to write the datum to the DRAM cell, and to read the datum that is stored in the DRAM cell. Because the datum is stored as a charge on a capacitor, it has a tendency to weaken over time due to a small amount of current that inexorably flows off of the capacitor as “leakage current”. For that reason, DRAM cells require periodic refreshing to restore, i.e. “recharge” the stored charge in the DRAM cell to a desirable level to maintain signal margin. However, not all DRAM cells of a DRAM can store a given amount of charge for the same time interval. Some DRAM cells which have relatively large leakage currents must be refreshed at more frequent intervals than other DRAM cells. It is the identification and characterization of such cells having larger leakage currents that needs to be performed better at time of production testing.
A particularly promising way of performing such testing of DRAM cells is by performing “write-window” testing. A “write-window” test seeks to identify and characterize marginal memory cells of a DRAM by varying the electrical operating conditions for reading and writing memory cells to a point which stresses the ability of the memory cells to retain stored data between successive refresh intervals.
In a “write-window” test, the time interval allotted to write a datum into a memory cell is shortened. This results in a smaller amount of charge being stored on the capacitor element of the memory cell than during normal operation. If the memory cell is then read after the refresh interval (or longer predetermined period of time) has passed, the faithfulness or lack thereof displayed by the memory cell in retaining the stored datum indicates whether the memory cell passes or fails during the stressed operating condition defined by the shortened writing interval.
In performing such “write-window” testing, it is desirable to test using a range of shortened writing intervals, in order obtain results having desirable granularity. Heretofore, it has not been possible to achieve a desired level of granularity when using available external test equipment to test advanced DRAM chips. Problems of attempting such testing are best illustrated with reference to the prior art organization of a DRAM shown in
A controller 122 generates many of the control signals needed to operate the DRAM including a column address strobe (CAS) signal which times the operation of the column decoder 118 and a row address strobe (RAS) signal which times the operation of the row decoder 110. Typically, the controller 122 also controls operation of the DRAM 100 to refresh the memory cells of the DRAM array 102 by providing a number “y” of control signal lines to the address control element 120 for controlling which addresses of the DRAM array 102 are presented to be refreshed, as well as the timing and control of the presentation of addresses for reading, writing refreshing the DRAM array 102. In addition, the controller 122 may also provide an autoprecharge timing signal (“APCHG”), which is timed in relation to the beginning of a DRAM read or write cycle (for example, as timed in relation to the RAS signal or WL_EN signal that the controller generates to time wordline decoding and driving operations). Such autoprecharge timing signal times the beginning of the bitline precharge interval to generally or closely coincide with the time that the accessed wordline is deactivated. For example, the autoprecharge timing signal is designed to begin the precharging of the bitlines at the same time that the accessed wordline is deactivated; i.e., “de-selected,” as by disabling the WL_EN signal. In addition to the autoprecharge timing signal, the controller 122 outputs a select (SEL) signal for selecting between an external precharge timing signal (PCHG) and the autoprecharge timing signal (APCHG) generated by the controller 122. Using the autoprecharge timing, the bitline precharging operation begins at a fixed interval of time after a particular wordline is activated using the WL_EN signal. By contrast, the external precharge signal (PCHG), arriving from external circuitry, e.g., external test equipment, is not so constrained. The external precharge signal PCHG can be timed to transition at various times, all depending upon the capability of the external test equipment and the fidelity of the signal channel from the test equipment into the DRAM 100 on the chip. However, there is a problem in that getting PCHG to transition at desired timings poses insurmountable challenges, as will become apparent in the following.
During normal operation, at time 152, the voltage on the wordline transitions from the low, inactive level to the high, active level. After the wordline is activated, the memory cells connected to that wordline can then be read, written or refreshed, the reading operation combining the latching and outputting the read data to SSA 116 (and on to the data I/O interface 114 (FIG. 1)), as well as refreshing all the memory cells connected to the wordline. At a later time 154, the wordline transitions again from the high voltage level to low voltage level. At that time, the memory cells are blocked from access, and the charge stored on the capacitors thereof is maintained until the next wordline access. This is followed by at least a minimum interval of time 156 required to precharge the bitlines of the DRAM 100 prior to the next activation of the same wordline. During such “precharge interval” 156, the voltages on non-accessed bitlines are reset to a predetermined voltage.
Thereafter, at a time 158, the DRAM is controlled to write a “0” to a memory cell. Preferably, the “0” is “strongly” written to the memory cell, i.e., written in a way which normally results in the voltage stored on the capacitor reaching the desired logic level. As shown in
Thereafter, at a later time 162 during operation, the memory cell is written to the logic “1”, i.e., the high level. However, this time, the writing operation is conducted using a shortened time interval for activating the wordline, ending at time 163, such that the “1” is only weakly written to the memory cell. Then, the accessed wordline is again deactivated. In order to test the memory cell under a fully stressed condition, the wordline is then maintained low for a period of time at least as long as the refresh interval tR, illustratively 128 ms, before being accessed again. The inactive period of time allows charge to escape from the capacitor of the memory cell via leakage current through any leakage paths which may be present. Thereafter, at time 164, the wordline is again activated and the contents of the memory cell are read therefrom.
At this time, when the memory cell tests good, the value of the datum read from the memory cell remains “1”, despite the stressed conditions under which the “1” is initially written to the memory cell, and the long interval of time that elapses before reading the memory cell again. However, if the memory cell is one that performs marginally or fails under normal conditions, the value in the memory cell might appear as a “0” instead when read. In such way, the marginal or failing memory cell is identified through the write-window test.
However, as noted above, one goal of testing is to determine with sufficient granularity the stressed conditions under which memory cells fail. Available external test equipment does not satisfy this goal. Curve 170 illustrates a waveform input to a time demultiplexer 124 of DRAM 100 (
Unfortunately, owing either to the external test equipment or the quality of the interface to the chip or both, the quality of the signal waveform 170 input to the DRAM 100 from the external test equipment is poor. This precludes this arrangement from providing adequate granularity in write-window testing. Pulse 172 of waveform 170 illustrates this point. The signal waveform arriving at demultiplexer 124 (
Accordingly, a new DRAM circuit and method are desired which are capable of providing write-window testing with greater granularity than that provided heretofore according to the prior art.
Therefore, according to an aspect of the invention, a method is provided for varying a length of an active duty cycle of a wordline of a semiconductor memory. In such method, a variable autoprecharge timing is set to a first value. Thereafter, the wordline is activated at a first point in time, after which the wordline is autoprecharged at a second point in time that is determined at least partly by the first value, wherein the length of the wordline active duty cycle is defined by the length of time between the first and second points in time.
According to another aspect of the invention, a method is provided for testing a memory cell within a semiconductor memory which has a test mode interface. Such method includes activating the test mode interface of the semiconductor memory, and thereafter, entering information through the test mode interface to set a variable autoprecharge timing to a first value. The autoprecharge timing is measured between a start of an activation interval of a wordline and a start of a precharge interval.
A bitline connected to the memory cell is then switched to a first logic level. An external precharge signal is provided to the semiconductor memory, and the wordline is activated to write the memory cell according to the first logic level, wherein an ending point of the active duty cycle of the wordline is determined at least partly by the external precharge signal.
Thereafter, a signal is provided to select the autoprecharge timing, and the bitline is switched to a second logic level. The wordline is activated to write the memory cell according to the second logic level, the length of the wordline activation interval being determined by the first value. The contents of the memory cell are then read to determine whether the memory cell successfully retains the second logic level.
According to another aspect of the invention, a semiconductor memory is provided which is operable in at least a test mode. Such semiconductor memory includes a memory array, the memory array including a plurality of memory cells, which are accessible through a plurality of respective wordlines and a plurality of respective bitlines. A row decoder driver is operable to activate one wordline of the plurality of wordlines at a first point in time that is determined in relation to a first signal and to deactivate the wordline at a second point in time that is determined in relation to a second signal;
The semiconductor memory further includes a precharge circuit which is operable to precharge the bitline at a third point in time, the third point in time occurring no sooner than the second point in time. A variable length delay circuit is operable to output the second signal at a delayed timing after the first signal, the delayed timing being delayed in relation to the first signal, the delayed timing having a controllably variable length.
For a more complete understanding of the present invention, and the advantages thereof, reference is now made to the following descriptions taken in conjunction with the accompanying drawing, in which:
Thus, a new DRAM circuit and method are provided herein which is capable of enabling “write-window” testing to be performed by varying the timing at which autoprecharging is performed in the DRAM. Since the beginning of the autoprecharging cycle is timed through internal operation of the DRAM circuit disclosed herein, it can be much more precisely controlled, and with much greater granularity than in the arrangement described above as background to the invention.
The embodiments of the invention will now be described with reference to
As further shown in
Illustratively, each selectable delay element 260 is implemented by a chain containing an even number of inverter devices, e.g., six inverters, fabricated on the same chip as the DRAM 200. These chains of inverter elements are used as granular selectable delay elements 260 in a series of such elements for generating the APCHG signal. In such way, a precisely controlled delay is achieved by passing the WL_EN signal through a selected portion of the series of selectable delay elements. In one embodiment, all selectable delay elements implement a delay for the same length of time. In another embodiment, the selectable delay elements implement delays corresponding to different lengths of time. For example, in one embodiment, some of the selectable delay elements produce a delay of one unit length of time, e.g., 0.5 ns, while other delay elements implement longer delay intervals, e.g., 1 ns, 2 ns, 3 ns, 4 ns, or longer delay intervals. In a particular embodiment, the delay elements implement only delay intervals, which vary by integer powers of an exponent. For example, the delay elements can implement delay intervals which vary only by integer powers of two, e.g., to implement a series of delays having delay intervals . . . 2-2 (¼), 2−1 (½), 20 (1), 21 (2), 22 (4), . . . By selecting delay elements having the same or different delay intervals, a total selected delay can be achieved by passing the signal through the series of selected delay elements. The switches 250 through 253 are double-acting switches, used to select whether the input to a particular one of the switches 250 through 253 is passed to a first output and on to the output of the PDD 232, after further isolation buffering and signal conditioning by isolation element 266 and latch 268. Alternatively, the input is passed on to the next selectable delay element 260 in the series of selectable delay elements.
A method of performing a “write-window” test according to an embodiment of the invention will now be described with reference to the flowchart depicted in
Thus, in block 310, test mode operation of the chip is entered by presenting the TM_EN signal to the test mode interface 230. The autoprecharge delay is then set (block 320) by TM_DATA provided to test mode interface 230, which, in turn, is passed on to the PDD 232 as the scan data input signal SD_IN. The scan data input to PDD 232 is timed in accordance with the scan data clock SD_CLK, and used to program a selected delay timing to activate the APCHG signal to begin the autoprecharging cycle. As indicated by block 330, after setting the autoprecharge delay interval, test mode operation is exited, in order to permit normal read, write and refreshing operations of the DRAM to be performed.
Thereafter, the controller 222 of the DRAM is used to perform a series of writing and reading operations, which together make up the write-window test, to test the ability of individual memory cells of the DRAM array to retain a stored datum over a given interval of time.
The operations depicted in flowchart are best understood with reference to the timing diagram provided in
Thereafter, at a time 458, coinciding with the next active edge 478 of WL_EN, the DRAM is controlled to write a “0” to a memory cell. Preferably, the “0” is “strongly” written to the memory cell, i.e., written in a way which normally results in the voltage stored on the capacitor reaching the desired logic level. As shown in
Thereafter, the autoprecharge timing feature of the DRAM 200 is enabled by presenting the AP_EN to the DRAM, which then selects the APCHG timing signal output by PDD 232 for controlling the SA/PCHG circuitry 254 of the DRAM. Accordingly, once AP_EN is active, the next rising edge 480 of the next WL_EN timing signal triggers a delayed signal APCHG which times the wordline to be deactivated and the bitlines to be precharged at an autoprecharge timing. The autoprecharge timing is determined in accordance with the programmed delay interval previously set as described above with reference to blocks 310 through 330 of
Accordingly, in block 360, this is done by setting the data I/O element 252 of the DRAM 200 to write a “1” to the DRAM and then activating the wordline (block 370) and maintaining the wordline active for only a shortened time interval tW. That time interval tW is determined by the autoprecharge timing that was set previously in accordance with the method described above (referring to blocks 310 through 330).
Then, in block 380 the accessed wordline is again deactivated when the active edge of the APCHG signal occurs. To test the memory cell under a fully stressed condition, the wordline is then maintained low for a period of time at least as long as the refresh interval tR, illustratively 128 ms, before being accessed again, the inactive period of time allowing charge to escape from the capacitor of the memory cell through any leakage paths which may be present.
Thereafter, at time 464, the wordline is activated again in accordance with the active edge 484 of the WL_EN signal, at which time the contents of the memory cell (block 390). At this time, when the memory cell passes test criteria, the value of the datum read from the memory cell remains “1”, despite the stressed conditions under which the “1” is initially written to the memory cell, and the long interval of time that elapses before reading the memory cell again. However, if the memory cell is one that performs marginally or fails under normal conditions, the value read from the memory cell can be “0” instead. In the case that the readout of the particular memory cell fails to produce a “1”, the address of the failing location of the memory cell is recorded, e.g., by outputting the current address over an external interface to equipment located off the chip. In addition, the conditions under which the memory cell fails are also recorded.
Desirably, all of the memory cells of the DRAM are tested in accordance with the method described above with reference to
Next, the interval of time tR is extended by which each of the wordlines of the array is accessed in turn after a designated retention period of time tR has elapsed. Such activation interval of wordlines is now the same as or optionally different from the standard refresh interval at which the DRAM is nominally refreshed. Thereafter, at time 464, triggered by the active edge 484 of the WL_EN signal, each memory cell of the DRAM is read out in the order that the memory cell was last written so as to test according to a uniform retention period for all memory cells of the chip.
The above method, performed using one autoprecharge delay timing, is desirably performed for different lengths of write-windows to characterize the DRAM. Each time, a different autoprecharge timing is set by presenting data and asserting the TM_EN signal at the test mode interface 230 (
Another benefit made possible by the invention will now be described. As indicated above, the write-window testing performed in accordance with
Although the invention herein has been described with reference to particular embodiments, it is to be understood that these embodiments are merely illustrative of the principles and applications of the present invention. It is therefore to be understood that numerous modifications may be made to the illustrative embodiments and that other arrangements may be devised without departing from the spirit and scope of the present invention as defined by the appended claims.
Patent | Priority | Assignee | Title |
7203127, | Sep 29 2005 | Polaris Innovations Limited | Apparatus and method for dynamically controlling data transfer in memory device |
8305835, | Dec 14 2010 | Advanced Micro Devices, Inc.; Advanced Micro Devices, INC | Memory elements having configurable access duty cycles and related operating methods |
9530463, | Apr 27 2010 | TAGMATECH, LLC | Memory device and method thereof |
9653149, | May 20 2015 | International Business Machines Corporation | Complementary bipolar SRAM |
9929161, | May 20 2015 | International Business Machines Corporation | Complementary bipolar SRAM |
Patent | Priority | Assignee | Title |
6272588, | May 30 1997 | SHENZHEN XINGUODU TECHNOLOGY CO , LTD | Method and apparatus for verifying and characterizing data retention time in a DRAM using built-in test circuitry |
6389564, | Jul 26 1999 | United Microelectronics Corp. | DRAM circuit having a testing unit and its testing method |
6453433, | Mar 30 1998 | Polaris Innovations Limited | Reduced signal test for dynamic random access memory |
6484278, | Aug 06 1998 | U S BANK NATIONAL ASSOCIATION, AS COLLATERAL AGENT | Method and apparatus for testing an embedded DRAM |
6512711, | Apr 25 1997 | Round Rock Research, LLC | Synchronous dynamic random access memory device |
6850454, | Jan 29 2003 | Renesas Technology Corp. | Semiconductor memory device with reduced current consumption during standby state |
20030185075, | |||
20040145959, |
Executed on | Assignor | Assignee | Conveyance | Frame | Reel | Doc |
Feb 02 2005 | Infineon Technologies AG | (assignment on the face of the patent) | / | |||
Feb 02 2005 | NIERLE, KLAUS | Infineon Technologies North America Corp | ASSIGNMENT OF ASSIGNORS INTEREST SEE DOCUMENT FOR DETAILS | 015816 | /0885 | |
Aug 29 2005 | Infineon Technologies North America Corp | Infineon Technologies AG | ASSIGNMENT OF ASSIGNORS INTEREST SEE DOCUMENT FOR DETAILS | 016464 | /0436 | |
Apr 25 2006 | Infineon Technologies AG | Qimonda AG | ASSIGNMENT OF ASSIGNORS INTEREST SEE DOCUMENT FOR DETAILS | 023821 | /0535 | |
Oct 09 2014 | Qimonda AG | Infineon Technologies AG | ASSIGNMENT OF ASSIGNORS INTEREST SEE DOCUMENT FOR DETAILS | 035623 | /0001 | |
Jul 08 2015 | Infineon Technologies AG | Polaris Innovations Limited | ASSIGNMENT OF ASSIGNORS INTEREST SEE DOCUMENT FOR DETAILS | 036873 | /0758 |
Date | Maintenance Fee Events |
Dec 31 2009 | M1551: Payment of Maintenance Fee, 4th Year, Large Entity. |
Dec 27 2013 | M1552: Payment of Maintenance Fee, 8th Year, Large Entity. |
Feb 12 2018 | REM: Maintenance Fee Reminder Mailed. |
Jul 30 2018 | EXP: Patent Expired for Failure to Pay Maintenance Fees. |
Date | Maintenance Schedule |
Jul 04 2009 | 4 years fee payment window open |
Jan 04 2010 | 6 months grace period start (w surcharge) |
Jul 04 2010 | patent expiry (for year 4) |
Jul 04 2012 | 2 years to revive unintentionally abandoned end. (for year 4) |
Jul 04 2013 | 8 years fee payment window open |
Jan 04 2014 | 6 months grace period start (w surcharge) |
Jul 04 2014 | patent expiry (for year 8) |
Jul 04 2016 | 2 years to revive unintentionally abandoned end. (for year 8) |
Jul 04 2017 | 12 years fee payment window open |
Jan 04 2018 | 6 months grace period start (w surcharge) |
Jul 04 2018 | patent expiry (for year 12) |
Jul 04 2020 | 2 years to revive unintentionally abandoned end. (for year 12) |